import type { EvalReport, TaskScore } from './EvalRunner.ts';
/** Mean judge score at or above which a non-assertion task counts as passed. */
export declare const DEFAULT_PASS_THRESHOLD = 0.7;
export interface BenchmarkTask {
    input: string;
    status: TaskScore['status'];
    /** true/false when judgeable, null for errored or unscored tasks. */
    passed: boolean | null;
    durationMs: number;
    tokens: number;
    scores: Record<string, number>;
}
export interface Benchmark {
    member: string;
    suite: string;
    generatedAt: string;
    provider?: string;
    model?: string;
    taskCount: number;
    evaluatedCount: number;
    errorCount: number;
    passRate: number | null;
    avgTimeMs: number | null;
    avgTokens: number | null;
    aggregate: Record<string, number>;
    tasks: BenchmarkTask[];
}
/**
 * Whether a task passes: an all-green assertion set wins; otherwise the mean of
 * its judge scores vs the threshold. Null when nothing is judgeable.
 */
export declare function taskPassed(task: TaskScore, threshold?: number): boolean | null;
/** Folds an EvalReport into a standardized benchmark summary. */
export declare function buildBenchmark(report: EvalReport, options?: {
    provider?: string;
    model?: string;
    threshold?: number;
}): Benchmark;
//# sourceMappingURL=benchmark.d.ts.map